Breast Cancer Tumor Detection Using Deep Learning Algorithms

Tumor classification is a crucial aspect of breast cancer diagnosis and treatment planning in medical imaging. Background: classification approaches often struggle with noise in samples, which reduces segmentation accuracy, affects feature selection, and hinders overall performance. Methods: a Graphical Neural Network (GNN)-based framework consisting of four key phases. First, images from the BreaKHis database are preprocessed using the Wiener filter to effectively reduce noise. Second, the watershed algorithm is applied for accurate image segmentation, enhancing the extraction of regions of interest. Third, Sequential Feature Selection (SFS) is utilized to identify optimal features without dimensionality loss. Finally, a GNN is employed for classification, leveraging its ability to model relationships between features and accurately classify tumors as benign or malignant. Achieving 98.26% accuracy, the proposed framework outperforms existing techniques, offering a robust and efficient tool for clinical breast cancer diagnostics.